Wondered what your take was on the fact that most of the AI-platforms we're talking about have been pre-trained on image-text pairs (i.e., photographs from lots of individual photographers) scraped from the Internet without permission or payment? That alone seems pretty unethical to me.
The biggest benefit you have is that you have been learning and doing fashion photography before AI came along. For the next generation, I wonder if they will see the reason to invest in large sensor cameras with exchangeable lenses when a lot can be simulated with some basic inputs. I do think part of fashion photography is safe - there's craftsmanship in the garment and it requires technical craftsmanship to capture it correctly too. And normally, a premium brand would want to have an authentic image of their work. Not something from AI. Even in the case of Levi's, they were mentioning that AI was to expand and supplement their range of models, not to replace all of them. At any rate - I hope we all grow with the tech.
In my opinion it may be a little early yet to speculate what the uses and limitations of generative AI may be. We can see some early utility for it, but not how it will ultimately shake out.
Advanced use of Stable Diffusion is a whole other world, it is indeed a bit behind Midjourney in terms of realism, stylization and coherence, but it's also compensated by a vast array of tools, it's truly amazing. Inpainting, ControlNet, photobashing, different models to choose from, training, etc. A lot of the problems you mentioned can be fixed with SD tools. It does take a bit of time to learn though. One way to use some of those advanced features without getting technical and investing a lot of time is using Leonardo AI, it's SD. By the way Midjourney has img2img feature, you might find it helpful, most people don't seem to know about it.
Thank you for pointing that out. It might be the push I need to get deeper into SD. The balance I'm looking for is a killer tool that can assist with a creative anomaly without requiring so much input control. Otherwise in setup time, it could be easier to just get in and photoshop something. But you are right that all the tools are going to develop and be opened up with more features. It's just a matter of time.
Good point T. Jay. Have you seen Luma AI? It turns videos into 3D models that can be rotated in any angle. It seems like a game changer. But yeah - still there's lots of human intervention to control and shoot the original material.
Great video and amazing results. It is unbelievable how AI created these realistic usable products and models! and the mixing between real photography and AI generated environment, it could save a lot of money for SMEs posting their products on social media or even website! Loved the video keep it up :)
Thanks for your insights, Eric! It's especially valuable that you didn't hide the fact that you photoshopped the logo. It seems like people tend to hide these things for some reason. I like to compare the current rise of AI with the development of digital sound in guitar music. People have split into two camps: some believe that the analog sound of a tube guitar amplifier is irreplaceable, and they struggle with all the downsides of that approach. Others are convinced that digital guitar processors have come so close to the analog tube sound of amplifiers that almost no one can tell the difference. Especially casual listeners rather than professional sound engineers. But the advantages and convenience of this approach are millions of times greater, starting from the price and ending with customization possibilities. I can see that in the future, people will have similar attitudes toward AI. For some, it will be a convenient but "artificial" tool. While others will hire photographers and have multi-day shoots, getting "analog" results in the old-fashioned way.
What a great analogy, Vlad! Have you seen that the AI generated song of Drake X The Weekend was banned from streaming? And at the same time the pop-star Grimes tweeted on Sunday to invite the public to use A.I. to create music in her voice. She said she would split 50% royalties on any successful AI generated song that uses her voice. There is so much happening right now its hard to stay on top!
cool video and thanks for sharing, the topic of this year.. many photographers feel threatened and rightfully so, but they too can make use of ai and create images they could never create before..
Such a valuable overview of the current state of AI. Thank you Eric! Have you found out anything about the legal permissions to use images like you've created here for client or publically published work? Because you generated them using AI do you own rights to the images? Or does the AI generator? Or someone else? I assume that's still a matter in limbo. I like how you noted image resolution as a small but significant limitation. Similarly, AI is not very helpful to us creators if we don't have the legal rights to use what we've generated. Grateful for any insight you can share!
It's a very good question. There are a couple of lawsuits going on at the moment. Although MidJourney says if you buy a corporate subscription plan, you have the rights to use the images in mass-media. But - in the US, AI-generated images are not protectable under current copyright law. It's a big grey area.
@@gro3a Ahh that's insightful, thanks! We shall all wait and see what gets decided. It feels very familiar to the Napster/iTunes/Spotify battle from back in the day. People started by illegal downloading free songs, Apply found a way to make it legal, Spotify found a way to make it cheap. Maybe this will play out similarly?
I have seen some sporting events that use AI cameras that are trained to follow the ball in football games. I can see that becoming a thing. I can see that also being used for TV talk shows and large events that require a static camera. I have also seen developments in drone AI cameras. These could be used to send out faster than a cameraperson for news events. We could see people hire drone cameras to photograph all kinds of things in reality. This technology will need more development than simply using the AI art generators. But it is completely possible for areas of image making where clients need realistic shots. I already follow a few channels that say their videos are edited by AI. So in theory we could get AI to cover most aspects of photography. Saying that in theory we could get AI to completely manage our farming and food production too. If AI is doing all the jobs, then we will have to think about why do we want to do anything. I got into photography to help me be more observant. I find going out with a camera I look at things more. Plus even though jobs get replaced doesn't stop people doing things. I know an artists that spent his life painting but never really sold them. He didn't care and can't be bothered with agents and galleries. He painted because it was something he wanted to do. Take money out of the equation and you refine your passion.
Yes - very cool with the AI controlled cameras for sports and getting different angles. I have seen plugins for video editing where it automatically cuts to the person talking during the podcast (with taste). It's very difficult to get a sense of the timeline of how fast this will all move. The tools already seem to be reinventing themselves every couple of weeks - despite being fresh.
You forgot a point , your misrepresenting both Jeep and North face , that roof rack does not exsit in the real world , same reason you can't do AI food photography its considered false advertisement.
It's illegal to show fake food in advertising, let alone rendered food, in most developed countries. Plus those food images looked disgusting. As for everything else, expect a serious consumer backlash when you don't hire "diverse" models and use computer generated fakes instead. Generally, people are starting to be disgusted by inequality, job loss, and tech mindless veneration.
What an amazing video, dude! Just posted a video about creative jobs that will be replaced by AI and showed a bit of this video there!
Very good statement, please produce more in this regards in details. Thank you
Next level work Eric. Thank you for generously sharing how you produced this for others to learn from your insights. Kind regards, Tom 🙇🏼♂
Wondered what your take was on the fact that most of the AI-platforms we're talking about have been pre-trained on image-text pairs (i.e., photographs from lots of individual photographers) scraped from the Internet without permission or payment? That alone seems pretty unethical to me.
Now using Photoshop is considered 'cheating', a great time to be alive hahah. Brilliant work btw Eric!
Great breakdown of the differences! I'm a fashion photographer... my job could be on the line or I can make the best of it and grow with the tech.
The biggest benefit you have is that you have been learning and doing fashion photography before AI came along. For the next generation, I wonder if they will see the reason to invest in large sensor cameras with exchangeable lenses when a lot can be simulated with some basic inputs. I do think part of fashion photography is safe - there's craftsmanship in the garment and it requires technical craftsmanship to capture it correctly too. And normally, a premium brand would want to have an authentic image of their work. Not something from AI. Even in the case of Levi's, they were mentioning that AI was to expand and supplement their range of models, not to replace all of them. At any rate - I hope we all grow with the tech.
In my opinion it may be a little early yet to speculate what the uses and limitations of generative AI may be. We can see some early utility for it, but not how it will ultimately shake out.
Great insights! Thanks for sharing it.
Superb video, thanks!
Thank you Eric! Excellent video and an inside of your creative process.
Amazing video keep it up man
Advanced use of Stable Diffusion is a whole other world, it is indeed a bit behind Midjourney in terms of realism, stylization and coherence, but it's also compensated by a vast array of tools, it's truly amazing. Inpainting, ControlNet, photobashing, different models to choose from, training, etc.
A lot of the problems you mentioned can be fixed with SD tools. It does take a bit of time to learn though. One way to use some of those advanced features without getting technical and investing a lot of time is using Leonardo AI, it's SD.
By the way Midjourney has img2img feature, you might find it helpful, most people don't seem to know about it.
Thank you for pointing that out. It might be the push I need to get deeper into SD. The balance I'm looking for is a killer tool that can assist with a creative anomaly without requiring so much input control. Otherwise in setup time, it could be easier to just get in and photoshop something. But you are right that all the tools are going to develop and be opened up with more features. It's just a matter of time.
@35:25 Add Drone/Aerial photography to the far right.
Good point T. Jay. Have you seen Luma AI? It turns videos into 3D models that can be rotated in any angle. It seems like a game changer. But yeah - still there's lots of human intervention to control and shoot the original material.
Great video and amazing results. It is unbelievable how AI created these realistic usable products and models! and the mixing between real photography and AI generated environment, it could save a lot of money for SMEs posting their products on social media or even website!
Loved the video keep it up :)
Thank you very much! New episode coming soon;) Stay tuned.
Thanks for your insights, Eric! It's especially valuable that you didn't hide the fact that you photoshopped the logo. It seems like people tend to hide these things for some reason.
I like to compare the current rise of AI with the development of digital sound in guitar music. People have split into two camps: some believe that the analog sound of a tube guitar amplifier is irreplaceable, and they struggle with all the downsides of that approach.
Others are convinced that digital guitar processors have come so close to the analog tube sound of amplifiers that almost no one can tell the difference. Especially casual listeners rather than professional sound engineers. But the advantages and convenience of this approach are millions of times greater, starting from the price and ending with customization possibilities.
I can see that in the future, people will have similar attitudes toward AI. For some, it will be a convenient but "artificial" tool. While others will hire photographers and have multi-day shoots, getting "analog" results in the old-fashioned way.
What a great analogy, Vlad! Have you seen that the AI generated song of Drake X The Weekend was banned from streaming? And at the same time the pop-star Grimes tweeted on Sunday to invite the public to use A.I. to create music in her voice. She said she would split 50% royalties on any successful AI generated song that uses her voice. There is so much happening right now its hard to stay on top!
good points!
cool video and thanks for sharing, the topic of this year.. many photographers feel threatened and rightfully so, but they too can make use of ai and create images they could never create before..
I think the learning for photographers is to own a style or subject matter that doesn't make sense for AI to do.
Such a valuable overview of the current state of AI. Thank you Eric! Have you found out anything about the legal permissions to use images like you've created here for client or publically published work? Because you generated them using AI do you own rights to the images? Or does the AI generator? Or someone else? I assume that's still a matter in limbo.
I like how you noted image resolution as a small but significant limitation. Similarly, AI is not very helpful to us creators if we don't have the legal rights to use what we've generated.
Grateful for any insight you can share!
It's a very good question. There are a couple of lawsuits going on at the moment. Although MidJourney says if you buy a corporate subscription plan, you have the rights to use the images in mass-media. But - in the US, AI-generated images are not protectable under current copyright law. It's a big grey area.
@@gro3a Ahh that's insightful, thanks! We shall all wait and see what gets decided. It feels very familiar to the Napster/iTunes/Spotify battle from back in the day. People started by illegal downloading free songs, Apply found a way to make it legal, Spotify found a way to make it cheap. Maybe this will play out similarly?
I have seen some sporting events that use AI cameras that are trained to follow the ball in football games. I can see that becoming a thing. I can see that also being used for TV talk shows and large events that require a static camera.
I have also seen developments in drone AI cameras. These could be used to send out faster than a cameraperson for news events. We could see people hire drone cameras to photograph all kinds of things in reality.
This technology will need more development than simply using the AI art generators. But it is completely possible for areas of image making where clients need realistic shots.
I already follow a few channels that say their videos are edited by AI. So in theory we could get AI to cover most aspects of photography.
Saying that in theory we could get AI to completely manage our farming and food production too. If AI is doing all the jobs, then we will have to think about why do we want to do anything. I got into photography to help me be more observant. I find going out with a camera I look at things more. Plus even though jobs get replaced doesn't stop people doing things. I know an artists that spent his life painting but never really sold them. He didn't care and can't be bothered with agents and galleries. He painted because it was something he wanted to do. Take money out of the equation and you refine your passion.
Yes - very cool with the AI controlled cameras for sports and getting different angles. I have seen plugins for video editing where it automatically cuts to the person talking during the podcast (with taste). It's very difficult to get a sense of the timeline of how fast this will all move. The tools already seem to be reinventing themselves every couple of weeks - despite being fresh.
@11:25 the logo doesn't say "The North Face"
All these photos use resources taken by photographers. It is injustice
You forgot a point , your misrepresenting both Jeep and North face , that roof rack does not exsit in the real world , same reason you can't do AI food photography its considered false advertisement.
It's illegal to show fake food in advertising, let alone rendered food, in most developed countries. Plus those food images looked disgusting. As for everything else, expect a serious consumer backlash when you don't hire "diverse" models and use computer generated fakes instead. Generally, people are starting to be disgusted by inequality, job loss, and tech mindless veneration.